LETTER doi:10.

1038/nature13314

A semi-synthetic organism with an expanded

genetic alphabet
Denis A. Malyshev1, Kirandeep Dhami1, Thomas Lavergne1, Tingjian Chen1, Nan Dai2, Jeremy M. Foster2, Ivan R. Corrêa Jr2
& Floyd E. Romesberg1

Organisms are defined by the information encoded in their genomes,
and since the origin of life this information has been encoded using a
two-base-pair genetic alphabet (A–T and G–C). In vitro, the alphabet
has been expanded to include several unnatural base pairs (UBPs)1–3.
We have developed a class of UBPs formed between nucleotides bear-
ing hydrophobic nucleobases, exemplified by the pair formed between
d5SICS and dNaM (d5SICS–dNaM), which is efficiently PCR-amplified1
and transcribed4,5 in vitro, and whose unique mechanism of repli-
cation has been characterized6,7. However, expansion of an organ-
ism’s genetic alphabet presents new and unprecedented challenges:
the unnatural nucleoside triphosphates must be available inside the
cell; endogenous polymerases must be able to use the unnatural tri-
phosphates to faithfully replicate DNA containing the UBP within
the complex cellular milieu; and finally, the UBP must be stable in
the presence of pathways that maintain the integrity of DNA. Here
we show that an exogenously expressed algal nucleotide triphosphate
transporter efficiently imports the triphosphates of both d5SICS and
dNaM (d5SICSTP and dNaMTP) into Escherichia coli, and that the
endogenous replication machinery uses them to accurately replicate
a plasmid containing d5SICS–dNaM. Neither the presence of the
unnatural triphosphates nor the replication of the UBP introduces a
notable growth burden. Lastly, we find that the UBP is not efficiently
excised by DNA repair pathways. Thus, the resulting bacterium is the
first organism to propagate stably an expanded genetic alphabet.
To make the unnatural triphosphates available inside the cell, we previ-
ously suggested using passive diffusion of the free nucleosides into the
cytoplasm followed by their conversion to the corresponding tripho-
sphate via the nucleoside salvage pathway8. Although we have shown
that analogues of d5SICS and dNaM are phosphorylated by the nucle-
oside kinase from Drosophila melanogaster8, monophosphate kinases
suggest that decomposition is mediated by phosphatases. As no degra-
dation was observed upon incubation in spent media, decomposition
seems to occur within the periplasm. No increase in stability was observed
in cultures of single-gene-deletion mutants of E. coli BW25113 lacking
a specific periplasmic phosphatase19 (as identified by the presence of a
Sec-type amino-terminal leader sequence), including phoA, ushA, appA,
aphA, yjjX, surE, yfbR, yjjG, yfaO, mutT, nagD, yggV, yrfG or ymfB,
suggesting that decomposition results from the activity of multiple phos-
phatases. However, the extracellular stability of [a-32P]-dATP was sig-
nificantly greater when 50 mM potassium phosphate (KPi) was added to
the growth medium (Extended Data Fig. 3). Thus, we measured [a-32P]-
dATP uptake from media containing 50 mM KPi after induction of the
transporter with isopropyl-b-D-thiogalactoside (IPTG) (Extended Data
Fig. 4). Although induction with 1 mM IPTG resulted in slower growth,
consistent with the previously reported toxicity of NTTs17, it also resulted
in maximal [a-32P]-dATP uptake. Thus, after addition of 1 mM IPTG,
we analysed the extracellular and intracellular stability of [a-32P]-dATP
as a function of time (Extended Data Fig. 5). Cells expressing PtNTT2
were found to have the highest levels of intracellular [a-32P]-dATP, and
although both extra- and intracellular dephosphorylation was still observed,
the ratio of triphosphate to dephosphorylation products inside the cell
remained roughly constant, indicating that the extracellular concen-
trations and PtNTT2-mediated influx are sufficient to compensate for
intracellular decomposition.
Likewise, we found that the addition of KPi increased the extracel-
lular stability of d5SICSTP and dNaMTP (Extended Data Fig. 2), and
a b
100 Media 3P
2P
Composition (%)

75 1P
0P
are more specific9, and in E. coli we found that overexpression of the
50
endogenous nucleoside diphosphate kinase results in poor growth. As O
N
O
an alternative, we focused on the nucleotide triphosphate transport- O S
O
O 25
ers (NTTs) of obligate intracellular bacteria and algal plastids10–14. We O d5SICS–dNaM O 0
expressed eight different NTTs in E. coli C41(DE3)15–17 and measured
the uptake of [a-32P]-dATP as a surrogate for the unnatural triphosphates 100 Cytoplasm

(Extended Data Fig. 1). We confirmed that [a-32P]-dATP is efficiently H

Composition (%)

N H 75
O
transported into cells by the NTTs from Phaeodactylum tricornutum N
O N O 50
(PtNTT2)18 and Thalassiosira pseudonana (TpNTT2)18. Although NTTs N
H N
N
O O
from Protochlamydia amoebophila (PamNTT2 and PamNTT5)15 also O H N
N 25
import [a-32P]-dATP, PtNTT2 showed the most activity, and both it O H O
0
and TpNTT2 are known to have broad specificity18, making them the dC–dG dA d5SICS dNaM
most promising NTTs for further characterization. [3P] (μM) 86 ± 9 30 ± 4
Transport via an NTT requires that the unnatural triphosphates are
sufficiently stable in culture media; however, preliminary characteriza- Figure 1 | Nucleoside triphosphate stability and import. a, Chemical
structure of the d5SICS–dNaM UBP compared to the natural dG–dC base pair.
tion of d5SICSTP and dNaMTP indicated that decomposition occurs
b, Composition analysis of d5SICS and dNaM in the media (top) and
in the presence of actively growing E. coli (Extended Data Fig. 2). Similar cytoplasmic (bottom) fractions of cells expressing PtNTT2 after 30 min
behaviour was observed with [a-32P]-dATP, and the dephosphorylation incubation; dA shown for comparison. 3P, 2P, 1P and 0P correspond to
products detected by thin-layer chromatography (TLC) for [a-32P]-dATP, triphosphate, diphosphate, monophosphate and nucleoside, respectively; [3P]
or by high-performance liquid chromatography (HPLC) and matrix- is the intracellular concentration of triphosphate. Error bars represent s.d.
assisted laser desorption/ionization (MALDI) for d5SICSTP and dNaMTP, of the mean, n 5 3.
1
Department of Chemistry, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, California 92037, USA. 2New England Biolabs, 240 County Road, Ipswich, Massachusetts 01938, USA.

1 5 M AY 2 0 1 4 | VO L 5 0 9 | N AT U R E | 3 8 5
©2014 Macmillan Publishers Limited. All rights reserved
 RESEARCH LETTER

when a stationary phase culture was diluted 100-fold into fresh media, for our purposes. Thirty minutes after their addition to the media, neither
the half-lives of both unnatural triphosphates (initial concentrations of of the unnatural triphosphates was detected in cells expressing TpNTT2;
0.25 mM) were found to be approximately 9 h, which seemed sufficient in contrast, 90 mM of d5SICSTP and 30 mM of dNaMTP were found in
a the cytoplasm of cells expressing PtNTT2 (Fig. 1b). Although intracel-
505
5'-CATGGTCATAGCTGTTTCCTGTGTGAAATTGTTATCCGCTCACA XTTCCACACAACATACGAGCCGGAAGC lular decomposition was still apparent, the intracellular concentrations
3'-GTACCAGTATCGACAAAGGACACACTTTAACAATAGGCGAGTGT YAAGGTGTGTTGTATGCTCGGCCTTCG
362 of intact triphosphate are significantly above the sub-micromolar KM
lacZα lac operator lac promoter
TK-1 Okazaki processing site
–10
values of the unnatural triphosphates for DNA polymerases20, setting
Primase recognition sequence the stage for replication of the UBP in a living bacterial cell.
0 The replication of DNA containing d5SICS-dNaM has been validated
5,00
in vitro with different polymerases, primarily family A polymerases, such
X
oter
lac op
era 500 Y as the Klenow fragment of E. coli DNA polymerase I (pol I)20,21. As the
om
I
lac

pr T7 promoter to
r Zα majority of the E. coli genome is replicated by pol III, we engineered a
Pt

cI
1,0

lac
NT

lac pro
la

00
T2

era
tor mo plasmid to focus replication of the UBP to pol I. Plasmid pINF (the infor-

1,0
op te
c

00
mation plasmid) was constructed from pUC19 using solid-phase DNA

r
la
4,000

pACS
5,151 bp synthesis and circular-extension PCR to replace the dA–dT pair at posi-
promoter

pINF
2,686 bp tion 505 with dNaM paired opposite an analogue of d5SICS (dTPT322)

o ri
2,500

(Fig. 2a, b). This positions the UBP 362 bp downstream of the ColE1
or

pR
Am
clo

pR
at

in origin of replication where leading-strand replication is mediated by

Am
DF

pro er m
m ote T7 t

0
0

50
pol I23, and within the TK-1 Okazaki processing site24, where lagging-
00

Am

1,
2,

pR
3,0
00 Sm
2,00
strand synthesis is also expected to be mediated by pol I. Synthetic
0
pINF was constructed using the d5SICS analogue because it should be
b e X
Y efficiently replaced by d5SICS if replication occurs in vivo, making it
pINF
X pUC19
possible to differentiate in vivo replicated pINF from synthetic pINF.
B To determine whether E. coli can use the imported unnatural triphos-
dX TP
PCR
dY′TP
PCR PCR dX TP phates to stably propagate pINF, C41(DE3) cells were first transformed
B dYTP
X
with a pCDF-1b plasmid encoding PtNTT2 (hereafter referred to as pACS,
+
X
Y′
Y
for accessory plasmid, Fig. 2a) and grown in media containing 0.25 mM
Plasmid pINF pUC19
of both unnatural triphosphates, 50 mM KPi and 1 mM IPTG to induce
IPTG + + – + + +
transporter production. Cells were then transformed with pINF, and after
dXTP/dYTP + + + – + –
PCR
dXTP
dY′TP SA – + + + + +
a 1-h recovery period, cultures were diluted tenfold with the same media
X 67% supplemented with ampicillin, and growth was monitored via culture
Y′
pINF
350 bp turbidity (Extended Data Table 1). As controls, cells were also transformed
200 bp
with pUC19, or grown without either IPTG or without the unnatural
100 bp
Transformation dXTP triphosphates. Again, growth was significantly slower in the presence
and growth dYTP
f X of IPTG, but the addition of d5SICSTP and dNaMTP resulted in only a
Y
X
Y pINF slight further decrease in growth in the absence of pINF, and interest-
pINF ingly, it eliminated a growth lag in the presence of pINF (Fig. 2c),
PCR
dXTP suggesting that the unnatural triphosphates are not toxic and are required
E. coli pACS dYTP
for the efficient replication of pINF.
X +
Y To demonstrate the replication of pINF, we recovered the plasmid
c dXTP from cells after 15 h of growth. The introduction of the UBP resulted in
0.6 dXTP IPTG
EP dYTP dYTP
+ –
+ + + + Figure 2 | Intracellular UBP replication. a, Structure of pACS and pINF.
– –
dX and dY correspond to dNaM and a d5SICS analogue22 that facilitated
0.4 plasmid construction (see Methods). cloDF, origin of replication;
pINF

– + Sm, streptomycin resistance gene; AmpR, ampicillin resistance gene; ori,

OD600

ColE1 origin of replication; lacZa, b-galactosidase fragment gene. b, Overview

of pINF construction. A DNA fragment containing the unnatural nucleotide
+ –
0.2 was synthesized via solid-phase DNA synthesis and then used to assemble
synthetic pINF via circular-extension PCR29. X, dNaM; Y9, dTPT3 (an analogue
+ +
of d5SICS22); y, d5SICS (see text). Colour indicates regions of homology. The
doubly nicked product was used directly to transform E. coli harbouring pACS.
pUC19

0
c, The addition of d5SICSTP and dNaMTP eliminates a growth lag of cells
11 13 15 17 19
Time (h) + – harbouring pINF. EP, electroporation. Error bars represent s.d. of the mean,
n 5 3. d, LC-MS/MS total ion chromatogram of global nucleoside content in
8
pINF and pUC19 recorded in dynamic multiple reaction monitoring (DMRM)
8 dXTP
d dC
IPTG
dYTP dG dA d5SICS mode. pINF and pUC19 (control) were propagated in E. coli in the presence or
+ + absence of unnatural triphosphates, and with or without PtNTT2 induction.
pINF

6
– +
Counts (×102)

6 + – The inset shows a 100-fold expansion of the mass-count axis in the d5SICS
region. e, Biotinylation only occurs in the presence of the UBP, the unnatural
pUC19

4
Counts (×104)

+ +
+ –
4
triphosphates and transporter induction. After growth, pINF was recovered,
2
and a 194-nucleotide region containing the site of UBP incorporation
(nucleotides 437–630) was amplified and biotinylated. B, biotin; SA,
2 0
dT 6.8 6.9 7.0 7.1
streptavidin. The natural pUC19 control plasmid was prepared identically to
pINF. A 50-bp DNA ladder is shown to the left. f, Sequencing analysis
0 demonstrates retention of the UBP. An abrupt termination in the Sanger
1.5 2.0 2.5 3.0 3.5 4.0 6.5 7.0 7.5 sequencing reaction indicates the presence of UBP incorporation (site indicated
Acquisition time (min) with arrow).

3 8 6 | N AT U R E | VO L 5 0 9 | 1 5 M AY 2 0 1 4
©2014 Macmillan Publishers Limited. All rights reserved

a small (approximately twofold) reduction in the copy number of pINF,
as gauged by its ratio to pACS (Extended Data Table 1); we determined
that the plasmid was amplified 2 3 107-fold during growth (approxi-
mately 24 doublings) based on the amount of recovered plasmid and
the transformation efficiency. To determine the level of UBP retention,
the recovered plasmid was digested, dephosphorylated to single nucle-
osides, and analysed by liquid chromatography-tandem mass spectrom-
etry (LC-MS/MS)25. Although the detection and quantification of dNaM
were precluded by its poor fragmentation efficiency and low product
ion counts over background, signal for d5SICS was clearly observable
(Fig. 2d). External calibration curves were constructed using the unnat-
ural nucleoside and validated by determining its ratio to dA in synthetic
oligonucleotides (Extended Data Table 2). Using the resulting calibra-
tion curve, we determined the ratio of dA to d5SICS in recovered pINF
was 1,106 to 1, which when compared to the expected ratio of 1,325
to 1, suggests the presence of approximately one UBP per plasmid. No
d5SICS was detected in control experiments in which the transporter
was not induced, or when the unnatural triphosphates were not added
to the media, or when pUC19 was used instead of pINF (Fig. 2d, inset),
demonstrating that its presence results from the replication of the UBP
and not from misinsertion of the unnatural triphosphates opposite a natural
nucleotide. Importantly, as the synthetic pINF contained an analogue
of d5SICS, and d5SICS was only provided as a triphosphate added to
the media, its presence in pINF confirms in vivo replication.
To independently confirm and quantify the retention of the UBP in
the recovered plasmid, the relevant region was amplified by PCR in the
presence of d5SICSTP and a biotinylated dNaMTP analogue4 (Fig. 2e).
Analysis by streptavidin gel shift showed that 67% of the amplified DNA
contained biotin. No shift was observed in control experiments where
the transporter was not induced, or when unnatural triphosphates were
not added, or when pUC19 was used instead of pINF, demonstrating
that the shift results from the presence of the UBP. Based on a calibra-
tion curve constructed from the shifts observed with the amplification
products of controlled mixtures of DNA containing dNaM or its fully
natural counterpart (Methods and Extended Data Fig. 6), the observed
gel shift corresponds to a UBP retention of 86%. Similarly, when the ampli-
fication product obtained with d5SICSTP and dNaMTP was analysed
by Sanger sequencing in the absence of the unnatural triphoshates1,26,27,
the sequencing chromatogram showed complete termination at the posi-
tion of UBP incorporation, which with an estimated lower limit of read-
through detection of 5%, suggests a level of UBP retention in excess of
95% (Fig. 2f). In contrast, amplification products obtained from pINF
recovered from cultures grown without PtNTT2 induction, without added
unnatural triphosphates, or obtained from pUC19 propagated under
identical conditions, showed no termination. Overall, the data unam-
biguously demonstrate that DNA containing the UBP was replicated
in vivo and allow us to estimate that replication occurred with fidelity
(retention per doubling) of at least 99.4% (24 doublings; 86% retention;
0.99424 5 0.86). This fidelity corresponds to an error rate of approxi-
mately 1023, which is comparable to the intrinsic error rate of some poly-
merases with natural DNA28.
The high retention of the UBP over a 15-h period of growth (approx-
imately 24 doublings) strongly suggests that it is not efficiently excised
by DNA repair pathways. To test further this hypothesis and to exam-
ine retention during prolonged stationary phase growth, we repeated
the experiments, but monitored UBP retention, cell growth and unnat-
ural triphosphate decomposition for up to 6 days without providing any
additional unnatural triphosphates (Fig. 3 and Extended Data Fig. 7). At
15 and 19 h of growth, the cultures reached an optical density at 600 nm
(OD600) of approximately 0.9 and 1.2, respectively, and both d5SICSTP
and dNaMTP decomposed to 17–20% and 10–16% of their initial 0.25-mM
concentrations (Extended Data Fig. 7a). In agreement with the experiments
described above, retention of the UBP after 15 h was 97 6 5% and .95%,
as determined by gel shift and sequencing, respectively, and after 19 h it
was 91 6 3% and .95%. As the cultures entered stationary phase and the
triphosphates decomposed completely, plasmid loss began to compete
©2014 Macmillan Publishers Limited. All rights reserved
LETTER RESEARCH
100
80 3
d5SICSTP (%)
Retention (%)
dNaMTP (%)
OD600
60
2
40
1
20
0 0
0 25 50 75 146
Time (h)
Figure 3 | Intracellular stability of the UBP. E. coli C41(DE3) pACS was
transformed with pINF and grown after a single dose of d5SICSTP and
dNaMTP was provided in the media. UBP retention in recovered pINF, OD600,
and relative amount of d5SICSTP and dNaMTP in the media (100% 5 0.25
mM), were determined as a function of time. Error bars represent s.d. of the
mean, n 5 3.
with replication (Extended Data Fig. 7b, c, d), but even then, retention
of the UBP remained at approximately 45% and 15%, at days 3 and 6
respectively. Moreover, when d5SICS-dNaM was lost, it was replaced
by dA–dT, which is consistent with the mutational spectrum of DNA
pol I20. Finally, the shape of the retention versus time curve mirrors that
of the growth versus time curve. Taken together, these data suggest that
in the absence of unnatural triphosphates, the UBP is eventually lost
by replication-mediated mispairing, and not from the activity of DNA
repair pathways.
We have demonstrated that PtNTT2 efficiently imports d5SICSTP
and dNaMTP into E. coli and that an endogenous polymerase, possibly
pol I, efficiently uses the unnatural triphosphates to replicate DNA con-
taining the UBP within the cellular environment with reasonable efficiency
and fidelity. Moreover, the UBP appears stable during both exponen-
tial and stationary phase growth despite the presence of all DNA repair
mechanisms. Remarkably, although expression of PtNTT2 results in a
somewhat reduced growth rate, neither the unnatural triphosphates nor
replication of the UBP results in significant further reduction in growth.
The resulting bacterium is the first organism that stably harbours DNA
containing three base pairs. In the future, this organism, or a variant with
the UBP incorporated at other episomal or chromosomal loci, should
provide a synthetic biology platform to orthogonally re-engineer cells,
with applications ranging from site-specific labelling of nucleic acids
in living cells to the construction of orthogonal transcription networks
and eventually the production and evolution of proteins with multiple,
different unnatural amino acids.
METHODS SUMMARY
To prepare electrocompetent C41(DE3) pACS cells, freshly transformed E. coli
C41(DE3) pACS was grown overnight in 2 3 YT medium (1.6% tryptone, 1% yeast
extract, 0.5% NaCl) supplemented with streptomycin and KPi. After 100-fold dilu-
tion into the same medium and outgrowth at 37 uC to OD600 5 0.20, IPTG was added
to induce expression of PtNTT2. After 40 min, cultures were rapidly cooled, washed
with sterile water and resuspended in 10% glycerol. An aliquot of electrocompetent
cells was mixed with pINF and electroporated. Pre-warmed 2 3 YT medium con-
taining streptomycin, IPTG and KPi was added, and an aliquot was diluted 3.3-fold
in the same media supplemented with 0.25 mM each of dNaMTP and d5SICSTP.
The resulting mixture was allowed to recover at 37 uC with shaking. After recovery,
cultures were centrifuged. Spent media was analysed for nucleotide composition by
HPLC (Extended Data Fig. 7a); cells were resuspended in fresh medium containing
streptomycin, ampicillin, IPTG, KPi and 0.25 mM each of dNaMTP and d5SICSTP,
and grown with shaking. At defined time points, OD600 was determined and ali-
quots were removed and centrifuged. Spent media were analysed for nucleotide
1 5 M AY 2 0 1 4 | VO L 5 0 9 | N AT U R E | 3 8 7
 RESEARCH LETTER
composition, and pINF was recovered by spin column purification. UBP retention
was characterized by LC-MS/MS, PCR amplification and gel electrophoresis, or
sequencing, as described in the Methods.
Online Content Any additional Methods, Extended Data display items and Source
Data are available in the online version of the paper; references unique to these
sections appear only in the online paper.
Received 27 November 2013; accepted 8 April 2014.
Published online 7 May 2014.
1. Malyshev, D. A. et al. Efficient and sequence-independent replication of DNA
containing a third base pair establishes a functional six-letter genetic alphabet.
Proc. Natl Acad. Sci. USA 109, 12005–12010 (2012).
2. Yang, Z., Chen, F., Alvarado, J. B. & Benner, S. A. Amplification, mutation, and
sequencing of a six-letter synthetic genetic system. J. Am. Chem. Soc. 133,
15105–15112 (2011).
3. Yamashige, R. et al. Highly specific unnatural base pair systems as a third base pair
for PCR amplification. Nucleic Acids Res. 40, 2793–2806 (2012).
4. Seo, Y. J., Malyshev, D. A., Lavergne, T., Ordoukhanian, P. & Romesberg, F. E.
Site-specific labeling of DNA and RNA using an efficiently replicated and
transcribed class of unnatural base pairs. J. Am. Chem. Soc. 133, 19878–19888
(2011).
5. Seo, Y. J., Matsuda, S. & Romesberg, F. E. Transcription of an expanded genetic
alphabet. J. Am. Chem. Soc. 131, 5046–5047 (2009).
6. Betz, K. et al. Structural insights into DNA replication without hydrogen bonds.
J. Am. Chem. Soc. 135, 18637–18643 (2013).
7. Betz, K. et al. KlenTaq polymerase replicates unnatural base pairs by inducing a
Watson–Crick geometry. Nature Chem. Biol. 8, 612–614 (2012).
8. Wu, Y., Fa, M., Tae, E. L., Schultz, P. G. & Romesberg, F. E. Enzymatic
phosphorylation of unnatural nucleosides. J. Am. Chem. Soc. 124, 14626–14630
(2002).
9. Yan, H. & Tsai, M. D. Nucleoside monophosphate kinases: structure, mechanism,
and substrate specificity. Adv. Enzymol. 73, 103–134 (1999).
10. Winkler, H. H. & Neuhaus, H. E. Non-mitochondrial ATP transport. Trends Biochem.
Sci. 24, 64–68 (1999).
11. Amiri, H., Karlberg, O. & Andersson, S. G. Deep origin of plastid/parasite ATP/ADP
translocases. J. Mol. Evol. 56, 137–150 (2003).
12. Hatch, T. P., Al-Hossainy, E. & Silverman, J. A. Adenine nucleotide and lysine
transport in Chlamydia psittaci. J. Bacteriol. 150, 662–670 (1982).
13. Winkler, H. H. Rickettsial permeability: an ADP-ATP transport system. J. Biol. Chem.
251, 389–396 (1976).
14. Horn, M. & Wagner, M. Bacterial endosymbionts of free-living amoebae. J. Eukaryot.
Microbiol. 51, 509–514 (2004).
15. Haferkamp, I. et al. Tapping the nucleotide pool of the host: novel nucleotide carrier
proteins of Protochlamydia amoebophila. Mol. Microbiol. 60, 1534–1545 (2006).
16. Miroux, B. & Walker, J. E. Over-production of proteins in Escherichia coli: mutant
hosts that allow synthesis of some membrane proteins and globular proteins at
high levels. J. Mol. Biol. 260, 289–298 (1996).
3 8 8 | N AT U R E | VO L 5 0 9 | 1 5 M AY 2 0 1 4
©2014 Macmillan Publishers Limited. All rights reserved
17. Haferkamp, I. & Linka, N. Functional expression and characterisation of membrane
transport proteins. Plant Biol. 14, 675–690 (2012).
18. Ast, M. et al. Diatom plastids depend on nucleotide import from the cytosol. Proc.
Natl Acad. Sci. USA 106, 3621–3626 (2009).
19. Baba, T. et al. Construction of Escherichia coli K-12 in-frame, single-gene knockout
mutants: the Keio collection. Mol. Syst. Biol. 2, 2006.0008 (2006).
20. Lavergne, T., Malyshev, D. A. & Romesberg, F. E. Major groove substituents and
polymerase recognition of a class of predominantly hydrophobic unnatural base
pairs. Chemistry 18, 1231–1239 (2012).
21. Seo, Y. J., Hwang, G. T., Ordoukhanian, P. & Romesberg, F. E. Optimization of an
unnatural base pair toward natural-like replication. J. Am. Chem. Soc. 131,
3246–3252 (2009).
22. Li, L. et al. Natural-like replication of an unnatural base pair for the expansion of the
genetic alphabet and biotechnology applications. J. Am. Chem. Soc. 136, 826–829
(2014).
23. Tomizawa, J. & Selzer, G. Initiation of DNA synthesis in Escherichia coli. Annu. Rev.
Biochem. 48, 999–1034 (1979).
24. Allen, J. M. et al. Roles of DNA polymerase I in leading and lagging-strand
replication defined by a high-resolution mutation footprint of ColE1 plasmid
replication. Nucleic Acids Res. 39, 7020–7033 (2011).
25. Hashimoto, H. et al. Structure of a Naegleria Tet-like dioxygenase in complex with
5-methylcytosine DNA. Nature 506, 391–395 (2013).
26. Malyshev, D. A., Seo, Y. J., Ordoukhanian, P. & Romesberg, F. E. PCR with an
expanded genetic alphabet. J. Am. Chem. Soc. 131, 14620–14621 (2009).
27. Hirao, I. et al. An unnatural hydrophobic base pair system: site-specific
incorporation of nucleotide analogs into DNA and RNA. Nature Methods 3,
729–735 (2006).
28. Goodman, M. F. Error-prone repair DNA polymerases in prokaryotes and
eukaryotes. Annu. Rev. Biochem. 71, 17–50 (2002).
29. Quan, J. & Tian, J. Circular polymerase extension cloning for high-throughput
cloning of complex and combinatorial DNA libraries. Nature Protocols 6, 242–251
(2011).
Supplementary Information is available in the online version of the paper.
Acknowledgements We thank I. Haferkamp and J. Audia for kindly providing the NTT
plasmids and helpful discussions, and P. Ordoukhanian for providing access to the
Center for Protein and Nucleic Acid Research at TSRI. This work was supported by the
US National Institutes of Health (NIH) (GM 060005).
Author Contributions D.A.M., K.D., T.C. and F.E.R. designed the experiments. D.A.M.,
K.D. and T.L. performed the experiments. N.D., J.M.F. and I.R.C.J. performed LC-MS/MS
analysis. D.A.M., K.D. and F.E.R. analysed data and D.A.M. and F.E.R. wrote the
manuscript with assistance from the other authors.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare competing financial interests: details
are available in the online version of the paper. Readers are welcome to comment on
the online version of the paper. Correspondence and requests for materials should be
addressed to F.E.R. (floyd@scripps.edu).

METHODS
Materials. 2 3 YT, 2 3 YT agar, IPTG, ampicillin and streptomycin were obtained
from Fisher Scientific. Ampicillin and streptomycin were used at 100 mg ml21 and
50 mg ml21, respectively. All pET-16b constructs containing the nucleotide trans-
porters were kindly provided by I. Haferkamp (Technische Universität Kaisers-
lautern, Germany) with the exception of pET16b-RpNTT2, which along with the
C41(DE3) E. coli strain, was provided by J. P. Audia (University of South Alabama,
USA). Plasmids pUC19 and pCDF-1b were obtained from Thermo Scientific and
EMD Millipore, respectively. Plasmids were purified using the PureLink Quick
Plasmid DNA Miniprep Kit (Life Technologies). OneTaq, DeepVent, Q5 Hot Start
High-Fidelity DNA Polymerases, and all restriction endonucleases were obtained
from New England Biolabs. In general, PCR reactions were divided into multiple
aliquots with one followed in real time using 0.5 3 Sybr Green I (Life Technologies);
following PCR, the aliquots were recombined, purified by spin column (DNA Clean
and Concentrator-5; Zymo Research, Irvine, California, USA) with elution in 20 ml
of water, then separated by agarose gel electrophoresis, followed by band excision
and recovery (Zymoclean Gel DNA Recovery Kit), eluting with 20 ml of water unless
stated otherwise. Polyacrylamide gels were stained with 1 3 Sybr Gold (Life Tech-
nologies) for 30 min, agarose gels were cast with 1 3 Sybr Gold. All gels were visu-
alized using a Molecular Imager Gel Doc XR1 equipped with 520DF30 filter (Bio-Rad)
and quantified with Quantity One software (Bio-Rad). The sequences of all DNA
oligonucleotides used in this study are provided in Supplementary Information.
Natural oligonucleotides were purchased from IDT (San Diego, California, USA).
The concentration of dsDNA was measured by fluorescent dye binding (Quant-iT
dsDNA HS Assay kit, Life Technologies) unless stated otherwise. The concentra-
tion of ssDNA was determined by UV absorption at 260 nm using a NanoDrop
1000 (Thermo Scientific). [a-32P]-dATP (25 mCi) was purchased from PerkinElmer
(Shelton, Connecticut, USA). Polyethyleneimine cellulose pre-coated Bakerflex TLC
plates (0.5 mm) were purchased from VWR. dNaM phosphoramidite, dNaM and
d5SICS nucleosides were obtained from Berry & Associates Inc. (Dexter, Michigan,
USA). Free nucleosides of dNaM and d5SICS (Berry & Associates) were converted to
the corresponding triphosphates under Ludwig conditions30. After purification by anion
exchange chromatography (DEAE Sephadex A-25) followed by reverse phase (C18)
HPLC and elution through a Dowex 50WX2-sodium column, both triphosphates
were lyophilized and kept at 220 uC until use. The d5SICSTP analogue dTPT3TP22
and the biotinylated dNaMTP analogue dmmo2SSBIOTP4 were made as reported
previously. MALDI-TOF mass spectrometry (Applied Biosystems Voyager DE-PRO
System 6008) was performed at the TSRI Center for Protein and Nucleic Acid Research.
Construction of NTT expression plasmids. The PtNTT2 gene was amplified from
plasmid pET-16b-PtNTT2 using primers PtNTT2-forward and PtNTT2-reverse;
the TpNTT2 gene was amplified from plasmid pET-16b-TpNTT2 using primers
TpNTT2-forward and TpNTT2-reverse. A linear fragment of pCDF-1b was gener-
ated using primers pCDF-1b-forward and pCDF-1b-reverse. All fragments were
purified as described in Materials. The pCDF-1b fragment (100 ng, 4.4 3 10214 mol)
and either the PtNTT2 (78 ng, 4.4 3 10214 mol) or TpNTT2 (85 ng, 4.4 3 10214 mol)
fragment were then assembled together using restriction-free circular polymerase
extension cloning29 in 1 3 OneTaq reaction buffer, MgSO4 adjusted to 3.0 mM,
0.2 mM of dNTP, and 0.02 U ml21 of OneTaq DNA under the following thermal
cycling conditions: initial denaturation (96 uC, 1 min); 10 cycles of denaturation
(96 uC, 30 s), touchdown annealing (54 uC to 49.5 uC for 30 s (20.5 uC per cycle)),
extension of 68 uC for 5 min, and final extension (68 uC, 5 min). Upon completion,
the samples were purified and used for heat-shock transformation of E. coli XL10.
Individual colonies were selected on lysogeny broth (LB)-agar containing strep-
tomycin, and assayed by colony PCR with primers PtNTT2-forward/reverse or
TpNTT2-forward/reverse. The presence of the NTT genes was confirmed by sequenc-
ing and double digestion with ApaI/EcoO109I restriction endonucleases with the
following expected pattern: pCDF-1b-PtNTT2 (2,546/2,605 bp), pCDF-1b-TpNTT2
(2,717/2,605 bp), pCDF-1b (1,016/2,605 bp). The complete nucleotide sequence
of the pCDF-1b-PtNTT2 plasmid (pACS) is provided in Supplementary Information.
Growth conditions to quantify nucleoside triphosphate uptake. E. coli C41(DE3)16
freshly transformed with pCDF-1b-PtNTT2 was grown in 2 3 YT with streptomy-
cin overnight, then diluted (1:100) into fresh 2 3 YT medium (1 ml of culture per
uptake with [a-32P]-dATP; 2 ml of culture per uptake with d5SICSTP or dNaMTP)
supplemented with 50 mM potassium phosphate (KPi) and streptomycin. A nega-
tive control with the inactive transporter pET-16b-RpNTT2, was treated identically
except ampicillin was used instead of streptomycin. Cells were grown to an OD600
of approximately 0.6 and the NTT expression was induced by the addition of IPTG
(1 mM). The culture was allowed to grow for another hour (final OD600 approxi-
mately 1.2) and then assayed directly for uptake as described below using a method
adapted from a previous paper15.
Preparation of media fraction for unnatural nucleoside triphosphate analysis.
The experiment was initiated by the addition of either dNaMTP or d5SICSTP
(10 mM each) directly to the media to a final concentration of 0.25 mM. Cells were
©2014 Macmillan Publishers Limited. All rights reserved
LETTER RESEARCH
incubated with the substrate with shaking at 37 uC for 30 min and then pelleted
(8,000 r.c.f. (relative centrifugal force) for 5 min, 4 uC). An aliquot of the media
fraction (40 ml) was mixed with acetonitrile (80 ml) to precipitate proteins31, and
then incubated at 22 uC for 30 min. Samples were either analysed immediately by
HPLC or stored at 280 uC until analysis. Analysis began with centrifugation (12,000
r.c.f. for 10 min at 22 uC), then the pellet was discarded, and the supernatant was
reduced to approximately 20 ml by SpeedVac, resuspended in buffer A (see below)
to a final volume of 50 ml, and analysed by HPLC (see below).
Preparation of cytoplasmic fraction for nucleoside triphosphate analysis. To
analyse the intracellular desphosphorylation of the unnatural nucleoside triphos-
phate, cell pellets were subjected to 3 3 100 ml washes of ice-cold KPi (50 mM).
Pellets were then resuspended in 250 ml of ice cold KPi (50 mM) and lysed with 250 ml
of lysis buffer L7 of the PureLink Quick Plasmid DNA Miniprep Kit (200 mM
NaOH, 1% w/v SDS), after which the resulting solution was incubated at 22 uC for
5 min. Precipitation buffer N4 (350 ml, 3.1 M potassium acetate, pH 5.5) was added,
and the sample was mixed to homogeneity. Following centrifugation (.12,000 r.c.f.
for 10 min, at 22 uC) the supernatant containing the unnatural nucleotides was
applied to a Hypersep C18 solid phase extraction column (Thermo Scientific) pre-
washed with acetonitrile (1 ml) and buffer A (1 ml, see HPLC protocol for buffer
composition). The column was then washed with buffer A and nucleotides were
eluted with 1 ml of 50% acetonitrile:50% triethylammonium bicarbonate (TEAB)
0.1 M (pH 7.5). The eluent was reduced to approximately 50 ml in a SpeedVac and
its volume was adjusted to 100 ml with buffer A before HPLC analysis.
HPLC protocol and nucleoside triphosphate quantification. Samples were applied
to a Phenomenex Jupiter LC column (3 mm C18 300 Å, 250 3 4.6 mm) and subjected
to a linear gradient of 0–40% B over 40 min at a flow rate of 1 ml min21. Buffer A:
95% 0.1 M TEAB, pH 7.5; 5% acetonitrile. Buffer B: 20% 0.1 M TEAB, pH 7.5; 80%
acetonitrile. Absorption was monitored at 230, 273, 288, 326 and 365 nm.
Each injection series included two extra control samples containing 5 nmol of
dNaMTP or d5SICSTP. The areas under the peaks that corresponded to tripho-
sphate, diphosphate, monophosphate and free nucleoside (confirmed by MALDI-
TOF) were integrated for both the control and the unknown samples (described
above). After peak integration, the ratio of the unknown peak to the control peak
adjusted for the loss from the extraction step (62% and 70% loss for dNaM and
d5SICS, respectively, Extended Data Table 3), provided a measure of the amount
of each of the moieties in the sample. To determine the relative concentrations of
unnatural nucleotide inside the cell, the amount of imported unnatural nucleotide
(dXTP, mmol) was then divided by the volume of cells, which was calculated as the
product of the volume of a single E. coli cell (1 mm3 based on a reported average value32;
that is, 1 3 1029 ml per cell) and the number of cells in each culture (OD600 of 1.0
equal to 1 3 109 cells per ml (ref. 32)). The RpNTT2 sample was used as a negative
control and its signal was subtracted to account for incomplete washing of nucle-
otide species from the media.
dATP uptake. To analyse the intracellular desphosphorylation of dATP, after induc-
tion of the transporter, the uptake reaction was initiated by the addition of dATP
(spiked with [a-32P]-dATP) to a final concentration of 0.25 mM, followed by incu-
bation at 37 uC with shaking for 30 min. The culture was then centrifuged (8,000
r.c.f. for 5 min at 22 uC). Supernatant was analysed by TLC. Cell pellets were washed
three times with ice-cold KPi (50 mM, 100 ml) to remove excess radioactive sub-
strate, lysed with NaOH (0.2 M, 100 ml) and centrifuged (10,000 r.c.f. for 5 min at
22 uC) to remove cell debris; supernatant was analysed by TLC.
TLC analysis. Samples (1 ml) were applied on a 0.5 mm polyethyleneimine cellu-
lose TLC plate and developed with sodium formate pH 3.0 (0.5 M, 30 s; 2.5 M, 2.5 min;
4.0 M, 40 min). Plates were dried using a heat gun and quantified by phosphorimaging
(Storm Imager, Molecular Dynamics) and Quantity One software.
Optimization of nucleotide extraction from cells for HPLC injection. To min-
imize the effect of the lysis and triphosphate extraction protocols on the decom-
position of nucleoside triphosphate within the cell, the extraction procedure was
optimized for the highest recovery with the lowest extent of decomposition (Extended
Data Table 3). To test different extraction methods, cells were grown as described
above, washed, and then 5 nmol of either dNaMTP or d5SICSTP was added to the
pellets, which were then subjected to different extraction protocols including boil-
ing water, hot ethanol, cold methanol, freeze and thaw, lysozyme, glass beads, NaOH,
trichloroacetic acid (TCA) with Freon, and perchloric acid (PCA) with KOH33.
The recovery and composition of the control was quantified by HPLC as described
above to determine the most effective procedure. Method 3—that is, cell lysis with
NaOH (Extended Data Table 3)—was found to be most effective and reprodu-
cible, thus we further optimized it by resuspension of the pellets in ice-cold KPi
(50 mM, 250 ml) before addition of NaOH to decrease dephosphorylation after cell
lysis (Method 4). Cell pellets were then processed as described above. See above for
the final extraction protocol.
Preparation of the unnatural insert for pINF construction. The TK-1-dNaM
oligonucleotide containing dNaM was prepared using solid-phase DNA synthesis
 RESEARCH LETTER
with ultra-mild DNA synthesis phosphoramidites on CPG ultramild supports
(1 mmol, Glen Research, Sterling, Virginia, USA) and an ABI Expedite 8905 syn-
thesizer. After the synthesis, the DMT-ON oligonucleotide was cleaved from the
solid support, deprotected and purified by Glen-Pak cartridge according to the
manufacturer’s recommendation (Glen Research), and then subjected to 8 M urea
8% PAGE. The gel was visualized by ultraviolet shadowing, the band corresponding
to the 75-mer was excised, and the DNA was recovered by crush and soak extraction,
filtration (0.45 mm), and final desalting over Sephadex G-25 (NAP-25 Columns, GE
Healthcare). The concentration of the single stranded oligonucleotide was deter-
mined by ultraviolet absorption at 260 nm assuming that the extinction coefficient of
dNaM at 260 nm is equal to that of dA. TK-1-dNaM (4 ng) was next amplified by
PCR under the following conditions: 1 3 OneTaq reaction buffer, MgSO4 adjusted
to 3.0 mM, 0.2 mM of dNTP, 0.1 mM of dNaMTP, 0.1 mM of the d5SICSTP ana-
logue dTPT3TP, 1 mM of each of the primers pUC19-fusion-forward and pUC19-
fusion-reverse, and 0.02 U ml21 of OneTaq DNA Polymerase (in a total of 4 3 50 ml
reactions) under the following thermal cycling conditions: initial denaturation (96 uC,
1 min) followed by 12 cycles of denaturation (96 uC, 10 s), annealing (60 uC, 15 s),
and extension (68 uC, 2 min). An identical PCR without the unnatural triphosphates
was run to obtain fully natural insert under identical conditions for the construction
of the natural control plasmid. Reactions were subjected to spin column purification
and then the desired PCR product (122 bp) was purified by a 4% agarose gel.
pUC19 linearization for pINF construction. pUC19 (20 ng) was amplified by
PCR under the following conditions: 1 3 Q5 reaction buffer, MgSO4 adjusted to
3.0 mM, 0.2 mM of dNTP, 1 mM of each primers pUC19-lin-forward and pUC19-
lin-reverse, and 0.02 U ml21 of Q5 Hot Start High-Fidelity DNA Polymerase (in a
total of 4 3 50 ml reactions with one reaction containing 0.5 3 Sybr Green I) under
the following thermal cycling conditions: initial denaturation (98 uC, 30 s); 20 cycles
of denaturation (98 uC, 10 s), annealing (60 uC, 15 s), and extension (72 uC, 2 min);
and final extension (72 uC, 5 min). The desired PCR product (2,611 bp) was purified
by a 2% agarose gel.
PCR assembly of pINF and the natural control plasmid. A linear fragment was
amplified from pUC19 using primers pUC19-lin-forward and pUC19-lin-reverse.
The resulting product (800 ng, 4.6 3 10213 mol) was combined with either the nat-
ural or unnatural insert (see above) (56 ng, 7.0 3 10213 mol) and assembled by cir-
cular overlap extension PCR under the following conditions: 1 3 OneTaq reaction
buffer, MgSO4 adjusted to 3.0 mM, 0.2 mM of dNTP, 0.1 mM of dNaMTP, 0.1 mM
of the d5SICSTP analogue dTPT3TP, and 0.02 U ml21 of OneTaq DNA Polymer-
ase (in a total of 4 3 50 ml reactions with one reaction containing 0.5 3 Sybr Green I)
using the following thermal cycling conditions: initial denaturation (96 uC, 1 min);
12 cycles of denaturation (96 uC, 30 s), annealing (62 uC, 1 min), and extension (68 uC,
5 min); final extension (68 uC, 5 min); and slow cooling (68 uC to 10 uC at a rate of
20.1 uC s21). The PCR product was analysed by restriction digestion on 1% aga-
rose and used directly for E. coli transformation. The d5SICS analogue dTPT322
pairs with dNaM, and dTPT3TP was used in place of d5SICSTP as DNA contain-
ing dTPT3–dNaM is better PCR amplified than DNA containing d5SICS–dNaM,
and this allowed for differentiation of synthetic and in vivo replicated pINF, as well
as facilitated the construction of high-quality pINF (UBP content .99%).
Preparation of electrocompetent cells for pINF replication in E. coli. C41(DE3)
cells were transformed by heat shock34 with 200 ng of pACS plasmid, and the trans-
formants were selected overnight on 2 3 YT-agar supplemented with streptomy-
cin. A single clone of freshly transformed C41(DE3) pACS was grown overnight in
2 3 YT medium (3 ml) supplemented with streptomycin and KPi (50 mM). After
100-fold dilution into the same fresh 2 3 YT media (300 ml), the cells were grown
at 37 uC until they reached an OD600 of 0.20 at which time IPTG was added to a
final concentration of 1 mM to induce the expression of PtNTT2. Cells were grown
for another 40 min and then growth was stopped by rapid cooling in ice water with
intensive shaking. After centrifugation in a prechilled centrifuge (2,400 r.c.f. for 10 min,
4 uC), the spent media was removed, and the cells were prepared for electropora-
tion by washing with ice-cold sterile water (3 3 150 ml). After washing, the cells
were resuspended in ice-cold 10% glycerol (1.5 ml) and split into 50-ml aliquots.
Although we found that dry ice yielded better results than liquid nitrogen for freezing
cells to store for later use, freshly prepared cells were used for all reported experi-
ments as they provided higher transformation efficiency of pINF and higher rep-
lication fidelity of the UBP.
Electroporation and recovery for pINF replication in E. coli. The aliquot of cells
was mixed with 2 ml of plasmid (400 ng), transferred to 0.2 cm gap electroporation
cuvette and electroporated using a Bio-Rad Gene Pulser according to the manu-
facturer’s recommendations (voltage 25 kV, capacitor 2.5 mF, resistor 200 V, time
constant 4.8 ms). Pre-warmed 2 3 YT media (0.95 ml, streptomycin, 1 mM IPTG,
50 mM KPi) was added, and after mixing, 45 ml was removed and combined with
105 ml of the same media (3.33-fold dilution) supplemented with 0.25 mM of dNaMTP
and d5SICSTP. The resulting mixture was allowed to recover for 1 h at 37 uC with
shaking (210 revolutions per min (r.p.m.)). The original transformation media
©2014 Macmillan Publishers Limited. All rights reserved
(10 ml) was spread onto 2 3 YT-agar containing streptomycin with 10- and 50-fold
dilutions for the determination of viable colony forming units after overnight growth
at 37 uC to calculate the number of the transformed pINF molecules (see the section
on calculation of the plasmid amplification). Transformation, recovery and growth
were carried out identically for the natural control plasmid. In addition, a negative
control was run and treated identically to pINF transformation except that it was
not subjected to electroporation (Extended Data Fig. 7b). No growth in the untrans-
formed negative control samples was observed even after 6 days. No PCR amp-
lification of the negative control was detected, which confirms that unamplified
pINF plasmid is not carried through cell growth and later detected erroneously as
the propagated plasmid.
Analysis of pINF replication in E. coli. After recovery, the cells were centrifuged
(4,000 r.c.f. for 5 min, 4 uC), and spent media (0.15 ml) was removed and analysed
for nucleotide composition by HPLC (Extended Data Fig. 7a). The cells were resus-
pended in fresh 2 3 YT media (1.5 ml, streptomycin, ampicillin, 1 mM IPTG, 50 mM
KPi, 0.25 mM dNaMTP, 0.25 mM d5SICSTP) and grown overnight at 37 uC while
shaking (250 r.p.m.), resulting in tenfold dilution compared to recovery media or
33.3-fold dilution compared to the originally transformed cells. Aliquots (100 ml)
were taken after 15, 19, 24, 32, 43, 53, 77 and 146 h, OD600 was determined, and the
cells were centrifuged (8,000 r.c.f. for 5 min, 4 uC). Spent media were analysed for
nucleotide composition by HPLC (Extended Data Fig. 7a), and the pINF and pACS
plasmid mixtures were recovered and linearized with NdeI restriction endonuclease;
pINF plasmid was purified by 1% agarose gel electrophoresis (Extended Data Fig. 7b)
and analysed by LC-MS/MS. The retention of the UBP on the pINF plasmid was
quantified by biotin gel shift mobility assay and sequencing as described below.
Mass spectrometry of pINF. Linearized pINF was digested to nucleosides by treat-
ment with a mixture of nuclease P1 (Sigma-Aldrich), shrimp alkaline phosphatase
(NEB), and DNase I (NEB), overnight at 37 uC, following a previously reported
protocol25. LC-MS/MS analysis was performed in duplicate by injecting 15 ng of
digested DNA on an Agilent 1290 UHPLC equipped with a G4212A diode array
detector and a 6490A Triple Quadrupole Mass Detector operating in the positive
electrospray ionization mode (1ESI). UHPLC was carried out using a Waters XSelect
HSS T3 XP column (2.1 3 100 mm, 2.5 mm) with the gradient mobile phase con-
sisting of methanol and 10 mM aqueous ammonium formate (pH 4.4). MS data
acquisition was performed in Dynamic Multiple Reaction Monitoring (DMRM)
mode. Each nucleoside was identified in the extracted chromatogram associated
with its specific MS/MS transition: dA at m/z 252R136, d5SICS at m/z 292R176,
and dNaM at m/z 275R171. External calibration curves with known amounts of
the natural and unnatural nucleosides were used to calculate the ratios of indivi-
dual nucleosides within the samples analysed. LC-MS/MS quantification was vali-
dated using synthetic oligonucleotides1 containing unnatural d5SICS and dNaM
(Extended Data Table 2).
DNA biotinylation by PCR to measure fidelity by gel shift mobility assay.
Purified mixtures of pINF and pACS plasmids (1 ng) from growth experiments
were amplified by PCR under the following conditions: 1 3 OneTaq reaction buf-
fer, MgSO4 adjusted to 3.0 mM, 0.3 mM of dNTP, 0.1 mM of the biotinylated
dNaMTP analogue dMMO2SSBIOTP, 0.1 mM of d5SICSTP, 1 mM of each of the
primers pUC19-seq-forward and pUC19-seq-reverse, 0.02 U ml21 of OneTaq DNA
Polymerase, and 0.0025 U ml21 of DeepVent DNA Polymerase in a total volume of
25 ml in an CFX Connect Real-Time PCR Detection System (Bio-Rad) under the
following thermal cycling conditions: initial denaturation (96 uC, 1 min); 10 cycles
of denaturation (96 uC, 30 s), annealing (64 uC, 30 s), and extension (68 uC, 4 min).
PCR products were purified, and the resulting biotinylated DNA duplexes (5 ml,
25–50 ng) were mixed with streptavidin (1 ml, 1 mg ml21, Promega) in phosphate
buffer (50 mM sodium phosphate, pH 7.5, 150 mM NaCl, 1 mM EDTA), incubated
for 30 min at 37 uC, mixed with 5 3 non-denaturing loading buffer (Qiagen), and
loaded onto 6% non-denaturing PAGE. After running at 110 V for 30 min, the gel
was visualized and quantified. The resulting fragment (194 bp) with primer regions
underlined and the unnatural nucleotide in bold (X represents dNaM or its bioti-
nylated analogue dMMO2SSBIO) is 59-GCAGGCATGCAAGCTTGGCGTAATC
ATGG TCATAGCTGTTTCCTGTGTGAAATTGTTATCCGCTCACAXTTCCA
CACAACATACGAGCCGGAAGCATAAAGTGTAAAGCCTGGGGTGCCTA
ATGAGTGAGCTAACTCACATTAATTGCGTTGCGCTCACTGCCCGCTTT
CCAGTCGGGAAACCTGTCGTGCCAG.
Streptavidin shift calibration for gel shift mobility assay. We have already reported
a calibration between streptavidin shift and the fraction of sequences with UBP in
the population (see Supplementary Fig. 8 of ref. 1). However, we found that spiking
the PCR reaction with DeepVent improves the fidelity with which DNA contain-
ing d5SICS-dMMO2SSBIO is amplified, and thus we repeated the calibration with
added DeepVent. To quantify the net retention of the UBP, nine defined mixtures
of the TK-1-dNaM template and its fully natural counterpart were prepared (Extended
Data Fig. 6a), subjected to biotinylation by PCR and analysed by mobility-shift assay
on 6% non-denaturing PAGE as described above. For calibration, the mixtures

TK-1-dNaM template and its fully natural counterpart with a known ratio of
unnatural and natural templates (0.04 ng) were amplified under the same condi-
tions over nine cycles of PCR with pUC19-fusion primers and analysed identically
to samples from the growth experiment (see the section on DNA biotinylation by
PCR). Each experiment was run in triplicate (a representative gel assay is shown in
Extended Data Fig. 6b), and the streptavidin shift (SAS, %) was plotted as function
of the UBP content (UBP, %). The data was then fit to a linear equation, SAS 5
0.77 3 UBP 1 2.0 (R2 5 0.999), where UBP corresponds to the retention of the
UBP (%) in the analysed samples after cellular replication and was calculated from
the SAS shift using the equation above.
Calculation of plasmid amplification. The cells were plated on 2 3 YT-agar
containing ampicillin and streptavidin directly after transformation with pINF,
and the colonies were counted after overnight growth at 37 uC. Assuming each cell
is only transformed with one molecule of plasmid, colony counts correspond to the
original amount of plasmid that was taken up by the cells. After overnight growth,
the plasmids were purified from a specific volume of the cell culture and quantified.
As purified plasmid DNA represents a mixture of the pINF and pACS plasmids,
digestion restriction analysis with NdeI exonuclease was performed to linearize
both plasmids, followed by 1% agarose gel electrophoresis (Extended Data Fig. 7b).
An example of calculations for the 19-h time point with one of three triplicates is
provided in Supplementary Information.
Fragment generation for Sanger sequencing to measure fidelity. Purified mix-
tures of pINF and pACS plasmids (1 ng) after the overnight growth were amplified
by PCR under the following conditions: 1 3 OneTaq reaction buffer, MgSO4 adjusted
to 3.0 mM, 0.2 mM of dNTP, 0.1 mM of dNaMTP, 0.1 mM of the d5SICSTP ana-
logue dTPT3TP, 1 mM of each of the primers pUC19-seq2-forward and pUC19-
seq-reverse (see below), and 0.02 U ml21 of OneTaq DNA Polymerase in a total
volume of 25 ml under the following thermal cycling conditions: initial denatura-
tion (96 uC, 1 min); and 10 cycles of denaturation (96 uC, 30 s), annealing (64 uC, 30 s),
and extension (68 uC, 2 min). Products were purified by spin column, quantified to
measure DNA concentration and then sequenced as described below. The sequenced
fragment (304 bp) with primer regions underlined and the unnatural nucleotide in
bold (X, dNaM) is 59-GCTGCAAGGCGATTAAGTTGGGTAACGCC AGGGT
TTTCCCAGTCACGACGTTGTAAAACGACGGCCAGTGAATTCGAGCTCG
GTACCCGGGGATCCTCTAGAGTCGACCTGCAGGCATGCAAGCTTGGCG
TAATCATGGTCATAGCTGTTTCCTGTGTGAAATTGTTATCCGCTCACAX
TTCCACACAACATACGAGCCGGAAGCATAAAGTGTAAAGCCTGGGGTG
CCTAATGAGTGAGCTAACTCACATTAATTGCGTTGCGCTCACTGCCCG
CTTTCCAGTCGGGAAACCTGTCGTGCCAG.
Sanger sequencing. The cycle sequencing reactions (10 ml) were performed on a
9800 Fast Thermal Cycler (Applied Biosystems) with the Cycle Sequencing Mix (0.5 ml)
of the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems) con-
taining 1 ng template and 6 pmol of sequencing primer pUC19-seq-reverse under
the following thermal cycling conditions: initial denaturation (98 uC, 1 min); and
25 cycles of denaturation (96 uC, 10 s), annealing (60 uC, 15 s), and extension (68 uC,
2.5 min). Upon completion, the residual dye terminators were removed from the
reaction with Agencourt CleanSEQ (Beckman-Coulter, Danvers, Massachusetts,
©2014 Macmillan Publishers Limited. All rights reserved
LETTER RESEARCH
USA). Products were eluted off the beads with deionized water and sequenced
directly on a 3730 DNA Analyzer (Applied Biosystems). Sequencing traces were
collected using Applied Biosystems Data Collection software v3.0 and analysed
with the Applied Biosystems Sequencing Analysis v5.2 software.
Analysis of Sanger sequencing traces. Sanger sequencing traces were analysed as
described previously1,26 to determine the retention of the unnatural base pair. In brief,
the presence of an unnatural nucleotide leads to a sharp termination of the sequencing
profile, whereas mutation to a natural nucleotide results in ‘read-through’. The extent
of this read-through after normalization is inversely correlated with the retention
of the unnatural base pair. Raw sequencing traces were analysed by first adjusting
the start and stop points for the Sequencing Analysis software (Applied Biosystems)
and then determining the average signal intensity individually for each channel (A,
C, G and T) for peaks within the defined points. This was done separately for the
parts of the sequencing trace before (section L) and after (section R) the unnatural
nucleotide. The R/L ratio after normalization (R/L)norm for sequencing decay and
read-through in the control unamplified sample (R/L 5 0.55(R/L)norm 1 7.2, see
ref. 26 for details) corresponds to the percentage of the natural sequences in the
pool. Therefore, an overall retention (F) of the incorporation of the unnatural base
pair during PCR is equal to 1 – (R/L)norm. As significant read-through (over 20%)
was observed in the direction of the pUC19-seq2-forward primer even with the
control plasmid (synthetic pINF); sequencing of only the opposite direction (pUC19-
seq-reverse) was used to gauge fidelity. Raw sequencing traces are shown in Fig. 2f
and provided as Supplementary Data.
30. Ludwig, J. & Eckstein, F. Rapid and efficient synthesis of nucleoside
59-0-(1-thiotriphosphates), 59-triphosphates and 2’,39-cyclophosphorothioates
using 2-chloro-4H-1,3,2-benzodioxaphosphorin-4-one. J. Org. Chem. 54,
631–635 (1989).
31. Alpert, A. & Shukla, A. Precipitation of Large, High-Abundance Proteins from
Serum with Organic Solvents in ABRF 2003: Translating Biology using Proteomics
and Functional Genomics Poster no. P111-W http://www.abrf.org/Other/
ABRFMeetings/ABRF2003/Alpert.pdf (2003).
32. Kubitschek, H. E. & Friske, J. A. Determination of bacterial cell volume with the
Coulter Counter. J. Bacteriol. 168, 1466–1467 (1986).
33. Yanes, O., Tautenhahn, R., Patti, G. J. & Siuzdak, G. Expanding coverage of the
metabolome for global metabolite profiling. Anal. Chem. 83, 2152–2161 (2011).
34. Seidman, C. E., Struhl, K., Sheen, J. & Jessen, T. Introduction of plasmid DNA into
cells. Curr. Prot. Mol. Biol. Chapter 1, Unit 1.8 (2001).
35. Knab, S., Mushak, T. M., Schmitz-Esser, S., Horn, M. & Haferkamp, I. Nucleotide
parasitism by Simkania negevensis (Chlamydiae). J. Bacteriol. 193, 225–235
(2011).
36. Audia, J. P. & Winkler, H. H. Study of the five Rickettsia prowazekii proteins
annotated as ATP/ADP translocases (Tlc): Only Tlc1 transports ATP/ADP, while
Tlc4 and Tlc5 transport other ribonucleotides. J. Bacteriol. 188, 6261–6268
(2006).
37. Hofer, A., Ekanem, J. T. & Thelander, L. Allosteric regulation of Trypanosoma brucei
ribonucleotide reductase studied in vitro and in vivo. J. Biol. Chem. 273,
34098–34104 (1998).
38. Reijenga, J. C., Wes, J. H. & van Dongen, C. A. M. Comparison of methanol
and perchloric acid extraction procedures for analysis of nucleotides
by isotachophoresis. J. Chromatogr. B Biomed. Sci. Appl. 374, 162–169 (1986).
 RESEARCH LETTER

Extended Data Figure 1 | Natural triphosphate uptake by NTTs. a, Survey of
reported substrate specificity (KM, mM) of the NTTs assayed in this study.
b, PtNTT2 is significantly more active in the uptake of [a-32P]-dATP compared
to other nucleotide transporters. Raw (left) and processed (right) data are
shown. Relative radioactivity corresponds to the total number of counts
produced by each sample. Interestingly, both PamNTT2 and PamNTT5 exhibit
a measurable uptake of dATP although this activity was not reported before.
This can possibly be explained by the fact that substrate specificity was only
characterized using competition experiments, and assay sensitivity might
not have been adequate to detect this activity15. References 35, 36 are cited in
this figure.

©2014 Macmillan Publishers Limited. All rights reserved

 LETTER RESEARCH

Extended Data Figure 2 | Degradation of unnatural triphosphates in down the dephosphorylation of both unnatural triphosphates.
growth media. Unnatural triphosphates (3P) of dNaM and d5SICS are a, Representative HPLC traces (for the region between ,20 and 24 min).
degraded to diphosphates (2P), monophosphates (1P) and nucleosides (0P) in dNaM and d5SICS nucleosides are eluted at approximately 40 min and not
the growing bacterial culture. Potassium phosphate (KPi) significantly slows shown. b, Composition profiles.

©2014 Macmillan Publishers Limited. All rights reserved

 RESEARCH LETTER

Extended Data Figure 3 | Effect of potassium phosphate on dATP uptake

and stability in growth media. a, KPi inhibits the uptake of [a-32P]-dATP at
concentrations above 100 mM. Raw (left) and processed (right) data are shown.
The NTT from Rickettsia prowazekii (RpNTT2) does not mediate the uptake of
any of the dNTPs and was used as a negative control: its background signal
was subtracted from those of PtNTT2 (black bars) and TpNTT2 (white bars).
Relative radioactivity corresponds to the total number of counts produced by
each sample. b, KPi (50 mM) significantly stabilizes [a-32P]-dATP in the
media. Triphosphate stability in the media is not significantly affected
by the nature of the NTT expressed. 3P, 2P and 1P correspond to triphosphate,
diphosphate and monophosphate states, respectively. Error bars represent
s.d. of the mean, n 5 3.

©2014 Macmillan Publishers Limited. All rights reserved

 LETTER RESEARCH

Extended Data Figure 4 | dATP uptake and growth of cells expressing
PtNTT2 as a function of inducer (IPTG) concentration. Growth curves and
[a-32P]-dATP uptake by bacterial cells transformed with pCDF-1b-PtNTT2
(pACS) plasmid as a function of IPTG concentration. a, Total uptake of
radioactive substrate (left) and total intracellular triphosphate content (right)
are shown at two different time points. Relative radioactivity corresponds to the
total number of counts produced by each sample. b, A stationary phase culture
of C41(DE3) pACS cells was diluted 100-fold into fresh 2 3 YT media
containing 50 mM KPi, streptomycin, and IPTG at the indicated
concentrations and were grown at 37uC. Error bars represent s.d. of the
mean, n 5 3.

©2014 Macmillan Publishers Limited. All rights reserved

 RESEARCH LETTER

Extended Data Figure 5 | Stability and uptake of dATP in the presence of respectively. 3P, 2P and 1P correspond to nucleoside triphosphate, diphosphate
50 mM KPi and 1 mM IPTG. Composition of [a-32P]-dATP in the media and monophosphate, respectively. M refers to a mixture of all three compounds
(left) and cytoplasmic fraction (right) as a function of time. TLC images and that was used as a TLC standard. The position labelled ‘Start’ corresponds to
their quantifications are shown at the bottom and the top of each of the panels, the position of sample spotting on the TLC plate.

©2014 Macmillan Publishers Limited. All rights reserved

 LETTER RESEARCH

Extended Data Figure 6 | Calibration of the streptavidin shift (SAS). a, The

SAS is plotted as a function of the fraction of template containing the UBP.
Error bars represent s.d. of the mean, n 5 3. b, Representative data.
SA, streptavidin.

©2014 Macmillan Publishers Limited. All rights reserved

 RESEARCH LETTER
Extended Data Figure 7 | Decomposition of unnatural triphosphates, pINF
quantification, and retention of the UBP with extended cell growth.
a, Dephosphorylation of the unnatural nucleoside triphosphate. 3P, 2P, 1P and
0P correspond to triphosphate, diphosphate, monophosphate and nucleoside
states, respectively. The composition at the end of the 1 h recovery is shown at
the right. b, Restriction analysis of pINF and pACS plasmids purified from
E. coli, linearized with NdeI restriction endonuclease and separated on a 1%
agarose gel (assembled from independent gel images). Molar ratios of pINF/
pACS plasmids are shown at the top of each lane. For each time point, triplicate
©2014 Macmillan Publishers Limited. All rights reserved
data are shown in three lanes with the untransformed control shown in the
fourth, rightmost lane (see Methods). c, Number of pINF doublings as a
function of time. The decrease starting at approximately 50 h is due to the loss
of the pINF plasmid that also results in increased error. See the section on pINF
replication in E. coli in the Methods for details. d, UBP retention (%) as a
function of growth as determined by gel shift (data shown in Fig. 3) and Sanger
sequencing (sequencing traces are available as Supplementary Data). In a, c
and d, error shown is the s.d. of mean, n 5 3.
 LETTER RESEARCH

Extended Data Table 1 | OD600 of E. coli cultures and relative copy

number of plasmid (pINF or control pUC19) as determined by its
molar ratio to pACS after 19 h of growth

X, NaM; Y, 5SICS.

©2014 Macmillan Publishers Limited. All rights reserved

 RESEARCH LETTER

Extended Data Table 2 | Relative quantification by LC-MS/MS using synthetic oligonucleotides containing d5SICS and dNaM

* dA/d5SICS and dA/dNaM ratios were calculated assuming that randomized nucleotides (N) around the unnatural base are distributed equally.

©2014 Macmillan Publishers Limited. All rights reserved

 LETTER RESEARCH

Extended Data Table 3 | Summary of the most successful extraction methods

* Recovery of all nucleotides (3P, 2P, 1P and nucleoside).

{ Calculated as a ratio of 3P composition (%) before and after the extraction.
References 37, 38 are cited in this figure. Details of methods 3 and 4 can be found online (http://2013.igem.org/wiki/images/e/ed/BGU_purelink_quick_plasmid_qrc.pdf).

©2014 Macmillan Publishers Limited. All rights reserved